Re: Considering replication w/mechanism for client-side updates

[Mark Meyers]

The problem with the first method is that a whole-DB copy occurs on each client every time an update occurs on any machine (when connected). When I say small, I think this DB will be 50-100 KB of data, but this method still means every update causes the master to sync to the slaves, and when it does, a whole DB copy operation occurs.

Another idea about handling this situation....

On the client side, I accumulate updates (in SQL) since the last connect time as a batch, such as in a file or db table. I also just run those updates as they happen on the client db.

When a connect occurs, and there are updates not yet sent up, then in Java I connect to the server, and send up the updates SQL.

Here is where my management has to come in. On the server, I connect to each client when available, and propagate updates to each client as needed, and each of them runs those updates on their copies.

So maybe this is my best option. It sounds like quite a bit more of my own coding to handle this method.